The problems were baked in.How it startedHow it started
You could say it all started with Elon Musk’s AI FOMO — and his crusade against “wokeness.” When his AI company, xAI, announced Grok in November 2023, it was described as a chatbot with “a rebellious streak” and the ability to “answer spicy questions that are rejected by most other AI systems.” The chatbot debuted after a few months of development and just two months of training, and the announcement highlighted that Grok would have real-time knowledge of the X platform.
But there are inherent risks to a chatbot having both the run of the internet and X, and it’s safe to say xAI may not have taken the necessary steps to address them. Since Musk took over Twitter in 2022 and renamed it X, he laid off 30% of its global trust and safety staff and cut its number of safety engineers by 80%, Australia’s online safety watchdog said last January. As for xAI, when Grok was released, it was unclear whether xAI had a safety team already in place. When Grok 4 was released in July, it took more than a month for the company to release a model card — a practice typically seen as an industry standard, which details safety tests and potential concerns. Two weeks after Grok 4’s release, an xAI employee wrote on X that he was hiring for xAI’s safety team and that they “urgently need strong engineers/researchers.” In response to a commenter, who asked, “xAI does safety?” the original employee said xAI was “working on it.”
...
pull down to refresh
related posts
archived link
The ambition to differentiate Grok by making it edgier and more willing to engage with controversial topics might appeal to a certain segment of users but it also raises clear concerns about accuracy harm prevention and trust. Real time integration with X creates both a potential strength and a risk since the platform itself has been under scrutiny for misinformation and reduced content moderation. Without a well established and adequately staffed safety team from the start xAI appears to have prioritized novelty over guardrails. The delayed release of the model card only reinforces the impression that governance protocols were secondary considerations. In AI as in any powerful technology safety cannot be an afterthought because it takes far less time to deploy than it does to repair the damage from a flawed launch. The real question is whether Elon Musk and xAI are prepared to invest the resources to build robust safeguards before they scale Grok further.
I don't know.
I'm read words: crusade, spicy questions, inherent risks, safety engineers, safety watchdog, safety team, safety tests.
Now, I've never used Grok before but I use other agents all day, every day. None of them seem remotely harmful. WTF is the problem?
Is the problem that it is training users to be national socialists as opposed to rainbow warriors, because, I'm pretty sure that if Grok trained users to ask about pronoun usage and holding hands together to defeat the evil that is carbon dioxide, I'd want some pretty stringent safety tests so my kids would not fall victim to the perils of socialist globalism.
They literally had not trained any safety towards nudity into Grok 4.
The AI-generated undressing has been bad - I know a few people that have gone from full Elmotard to eternally disgusted over it. I think the problem is moving fast and breaking things on a platform that has a couple 100M daily users.
Okay. Thanks for the upshot. I read more and got a better idea.
Thing is like the article states, any of this can be done, has been done, .. and ultimately will be done. If it wasn't Grok, it'd be ChatGPT, I guess. I suppose the only difference is that it doen't have the reach.
Pretty sure, it wouldn't be so different 'share image' 'upload to facebook'. Something seems off to me. There's undoubtedly a campaign against the 'free speech media'.
I feel at a loss, whenever I hear of people self-harming because of ChatGPT, or indecent deep-faking on x. Probably because I don't use eith, never have, probably never will. Besides I don't share images of myself. If I was a celebrity, maybe I'd be a bit pissed over CGI rendering of myself. But am I even allowed to say it just sounds so overblown?
There are probably a million apps out there that will let you nudify whatever you want. The difference is indeed reach [1] and that's why I don't like what happened to Twitter in the past couple of years (also before Elon took over fwiw.) Good stewardship means you do not alienate your users with mistakes, as established products aren't a good sandbox. Dividing social networks into political bubbles is awful, imho; all it does is create echo chambers.
I do think that there's a lot of hysteria out there, in this I'm totally with you. In this case though, it's been pretty bad in terms of not only the xAI team not doing good work up front, but also being pretty awful on the response.
Caught with their pants down, no stable diffusion needed.
This is also why on my "dayjob" FOSS project I maintain I'm a complete a-hole when someone comes in with some bullshit PR and tries to push it instead of listening to review comments: I have to defend the installed base (millions) that are used to very high quality code. I don't care if someone forks that repo and does their own thing (as long as they don't lie or ride my years of careful maintenance work as a reason why their stuff is great) but it's not going into the main repo if it sucks, and whenever I feel my own code isn't properly reviewed, I force a hold off merge on that too (which happens a lot these past 2 years.) ↩
re: 1
That gave me flashbacks to last year's Core PR/force push tribal vibes.
I have no more respect to give to anyone than I have for FOSS developers, though I don't donate as much as I should to the maintainers of software I use.
I guess the ethos of the handfull of projects that serve the 90+% of ai and social platforms are not necessarily as conscientious.
Interesting, that where one platform alienates its userbase, another one gains traction.
this ^ and 👇
Threads overtakes X on mobile, but still lags far behind
Threads edges out X in daily mobile users, new data shows